* Render attachments inside iframes.
* Fix up the image and video views
* Fix m.audio
* Comments, and only use the cross domain renderer if the attachment is encrypted
* Fix whitespace
* Don't decrypt file attachments immediately
* Use https://usercontent.riot.im/v1.html by default
* typos
* Put the config in the React context.
Use it in MFileBody to configure the cross origin renderer URL.
* Call it appConfig in the context
* Return the promises so they don't get dropped
`onWidgetLoad` is now being called when an image has been decrypted so that the ScrollPanel maintains its scroll position (whether it's stuckAtBottom or not). This attempts to fix https://github.com/vector-im/riot-web/issues/2624
This increases `UNPAGINATION_PADDING` (see the ASCII on ScrollPanel.js, `_getExcessHeight`), and also debounces unfilling requests made for 200ms. This forces unfilling requests not to be sent unless the next 200ms has no scrolling, effectively.
Fix scroll up, down pagination test
NB: this test may not fail on Travis, although it did fail locally without a fix: #563.
Once the test has scrolled the panel to the top, to the earliest events, it should be able to forward paginate, because some degree of unpagination occurs. This does assume that unpagination will occur when scrolling to the beginning of the events and that unpagination should allow pagination again in the same direction.
Instead of checking that the first event is no longer the first event (varies due to unpagination), check instead that the most recent event can be seen when scrolling all the way down to the bottom of the TimelinePanel.
Scrolling past the bottom of content seems to have strange behaviour, which isn't a useful part of the test. So now the test will scroll down until the last event instead.
Fix pagination issue where recent events are lost
Scrolling up a few pages followed by scrolling down to the most recent events previously caused some events to go missing. A test has been modified in conjunction with this fix to make sure that this failure mode is tested for in future. This commit should fix the issue, and the most recent events should be paginated back in.
Also, the net change of nil is detected as having the first and last events being _different_. The summary should only include those that have their first and last events being the _same_ because that is a net change (within the block of member events).
